Yield Forecasting Models based on Weather Parameters for Eastern U.P.

P; ey KK; Maurya D; Gupta G; Mishra SV

doi:10.5958/2229-4473.2016.00006.9

Research Article, Vegetos Vol: 29 Issue: 1

Yield Forecasting Models based on Weather Parameters for Eastern U.P.

*Pandey KK^1, Maurya D², Gupta G² and Mishra SV³**
¹Department of Agriculture Station, S K College of Agriculuture & Research Station, Kawardha, Chhattisgarh, India
²Department of Agriculture and Allied Sciences, Baba Farid Institute of Technology, Suddhowala, Dehradun, India
³Department of Agriculture Physics, Indian Agriculture Research Institute, New Delhi, India
Corresponding author : Pandey KK Department of Agriculture Stat., S K College of Agriculture & Research Station, Kawardha, Chhattisgarh, India Tel: 07741233300 E-mail: kkpandeystat@gmail.com
Received: November 13, 2015 Accepted: February 16, 2016 Published: February 23, 2016
Citation: Pandey KK, Maurya D, Gupta G, Mishra SV (2016) Yield Forecasting Models based on Weather Parameters for Eastern U.P. Vegetos 29:1. doi: 10.5958/2229-4473.2016.00006.9

Abstract

Yield Forecasting Models based on Weather Parameters for Eastern U.P.

In the present study, attempts have been made to development of pre harvest models for pre harvesting forecasting of rice yield at district level on the basis of generated weather variables. Weekly data (14 meteorological weeks) because Time of the pre harvest forecast is 14th week 2½ month before harvesting flowering stage and seven weather variables over a span of 20 years period (1995 to 2014) has been used along with the annual rice production data for respective year for eastern UP. The data of de-trend yield and generated weather variables for 18 years of has been used for generation of the models. (1995-2012). Total five models were validated with 2 year independent data set (2013 and 2014). The (PPE) ranging - 4.32 to 10.56 % from the observed yield for all 5 models. Highest correlation has been found 0.68 between Z231 (Weighted interaction between MaxT and MinT). Significance of models are obtained by on the basis of highest R2 (86%) and by (P-value) (0.08). MBE (Mean Biased Error), RMSE (Root Mean Square Error) and PE (Percent Error) has been used for the Error analysis. On the basis of R2 and other parameter the model IV has been found best..

Keywords: Correlation coefficient; MBE; PE; Rice; RMSE

Download PDF

Keywords

Correlation coefficient; MBE; PE; Rice; RMSE

Introduction

Rice (Oryza sativa), one of the three most important food grain crops in world, forms the staple diet of 2.7 billon people. The effect of weather parameters at different stages of growth of crop may help in understanding their response in term of final yield and also provide a forecast of crop yields in advance before the harvest. Changes in the timing of phenological events are among the most important indicators of global warming reported, phonological change due to increasing of temperature. The forecasting equations have also been developed for forecasting paddy yield Shankar and Gupta [1] and for wheat yield for Kanpur district U.P. Agarwal et al. [2].

The majority of the agriculture land is used to grow major cereal crops; rice and wheat. Rice is the major crop in Uttar Pradesh and is grown in about 5.90 Mha which comprises of 13.5% of total rice in India. Uttar Pradesh has favourable and suitable climate, vast areas of fertile soils, sunshine and adequate water resources. The cropping intensity is 153% of the Uttar Pradesh. Uttar Pradesh ranks 3rd in the country in production of rice [3].

Agriculture is the main source of income for families in the Uttar Pradesh and India as well. It has 11.56 million hectare of cultivated area, constituting 70% of the total geographical area. The irrigated area is over 13.43 million hectare. The small and marginal farmers jointly contribute 19.46% of farming household in eastern region against that of 19.11% of Uttar Pradesh. Agriculture is the most important in the India because about 65% of its population resides in rural areas and 70% of the total workers are involved directly or indirectly in cultivation/farming which accounts for 27% of Uttar Pradesh GDP.

Agriculture is an economic sector which depends highly on climatic conditions. Crop models are frequently used to evaluate the ability of climate forecasts in guiding crop management practices. In statistical model approach, one or several variables (representing weather) are related to crop responses such as yield and yield contributing characters. Crop yield is the integrated effect of a number of physical and physiological processes that occur during the crop-growing period. These processes are influenced by the characteristics of the crop, weather, soil and time management factors [4]. Statistical models provide simple alternative to process- based models. The main advantages of statistical models are their limited reliance on field calibration data and their transparent assessment of model uncertainties [5]. Rice is cultivated during Kharif season (June to October- November) at the area of eastern Uttar Pradesh.

In the present study, an attempt has been made to develop suitable statistical models for forecasting of pre-harvest crop yield in faizabad district from the weekly data on weather variables with a few modifications.

Materials and Methods

Weekly weather data of crop growing season for 20 years (1995-2014) of Faizabad districts (260 47’ N latitude and 820 12’E longitudes and 113 m above mean sea level) has been collected from the Meteorological Station of department of Agro-meteorology N. D. University of Agriculture & Technology, Faizabad Uttar Pradesh, India and 14 week data has been used for pre harvest forecast modelling under the present study. The data on 14 weeks are used because that is flowering stage of the rice crop. The yield data of the rice crop for Faizabad district of Uttar Pradesh for 20 years (1995-2014) have been obtained from Directorate of Agricultural Statistics and Crop Insurance, Government of Uttar Pradesh, U.P. Weekly data of give weather variables rainfall (mm), maximum temperature (maxT)°C, minimum temperature (minT)°C, morning relative humidity (RH-I) %, afternoon relative humidity (RH-II) % has employed according to growing period of crop. Correlations were worked out between weather parameters (independent variable) with respective year and yield of crop. The yield data was de-trended.

De-trend Yield

Y= a + bt

Where; Y, a, b and t is observed yield, constant, regression coefficient and time trend respectively. Transformation of weekly weather data into new set. Two new variables from each weather variables (consisting of 14 meteorological weeks) have been generated as follows:

Let X_iw be the value of i^th (i = 1, 2,…,5) weather variable at wth weeks (w = 1,2,…..m) in this study m is 14.

The unweighted generated variables have been generated as follows:

i- denotes weather variable (i=1 to 5)

w- denotes week number (w= 1 to 14) during the growth of rice crop.

The weighted generated variables were computed as follows:

Let r_iw be the simple correlation coefficient between weather variable X_iw at w^th week and crop yield over the period of 18 years used as a weight. The weighted variables have been generated as follows [6]:

Thus we now have a total of 10 weather variables consisting of 5 weighted and 5 unweighted variables. The following models are then fitted to study the effect of weather variables:

Effect introduction of weather variables on crop yield

For studying the joint effect of two weather variables on crop-yield has been extended by including interaction terms in the model as follows:

where r_ii’,_w is the correlation coefficient between crop yield (detrended) Y and product of weather variables X_iw and X_i’wclearly, we have two generated variables (interaction term)

the unweighted one.

and

the weighted one.

Model fitting

Five models were attempted, one for the set of unweighted generated variables, second for the set of weighted generated weather variables, third for the unweighted + weighted generated variables taken all the 10 variables together simultaneously, fourth for the unweighted interaction weather variables and fifth for the weighted interaction weather variables; used as independent variable and yield as dependent.

Model-I	Y_i=a + b_iZ_i(unwt) + ε_i	Model-II	Y_i =a + b_iZ_i(wt) + ε_i
Model-III	Y_i =a + b_iZ_i( inter unwt) + ε_i	Model-IV	Y_i =a + b_iZ_i( inter wt) + ε_i
Model-V	Y_i=a + b_iZ_i(unwt) + b_iZ_i(wt) + ε_i		i=1,2,……,5

Where Y_i is yield, a is generalized constant ai’s (i=1 to 5) are model parameter associated with unweighted weather variables, b_i’s are model parameters associated with weighted weather variables and ε_i is error term supposed to follow normal distribution with mean zero and variance σ².

Statistical Analysis of Data

The statistical analysis of data was performed using SPSS-20 and MS Excel used for estimating the model parameters using multiple regression procedures. The validation of the model was also performed using by the Percent Prediction Error (PPE), Mean Bias Error (MBE) and Root Mean Square Error (RMSE) [7].

Percent prediction error (PPE)

Mean bias error (MBE)

Root mean square error (RMSE)

Result and Discussion

Positive correlation 0.68 found between yield and interaction of Maximum & Minimum Temperature and same result found in yield and weighted Rainfall, minimum and negative correlation -0.47 found between yield & un weighted Minimum Temperature, followed by yield and interaction un weighted Minimum Temperature & Morning Relative humidity (-0.40) (Table 1). Highest R² has been found in model IV followed by model III i.e. 86 % and 81 % respectively (Table 2). Model IV found most significant 0.08 followed by model I. Singh et al. [8] reported for similar results on the basis of temperature and rainfall on wheat yield at south western region of Punjab

Table 1: Correlation between generated weather variables and yield.

Table 2: Statistical models for rice yield forecast at Faizabad district.

The observed and forecasted yields for period (2013-2014) have been presented in Table 3 and various error analysis of independent and all data set in Table 4. The regressions models were validated with the two years (2013-2014) of independent data set. The data exposed that the models deviated (PPE) -4.32 to 10.56 % from the observed yield; the error analysis revealed that the MBE, RMSE and PE of models are ranging between -22.81 kg ha^-1 to – 6.56 kg ha^-1, RMSE 175.84 to 205.25 kg ha^-1, and 6.81 % to 7.97 % respectively for all data set and 59.07 to 205.32 kg ha^-1, 65.57 to 205.40 kg ha-1 and 0.49% to 7.40% respectively for independent data set. Similar results have also been reported by Kumar et al. [9] for the rice wheat and sugarcane under the Navsari and Bharuch districts. Pandey et al. [10] also reported the pre harvest forecast models for the rice crop for faizabad district of eastern U.P.

Table 3: Validation of models.

Table 4: Error analysis of forecasting models.

The study is based on the generated weather variables on the collected weather data (raw data) [11]. This is very important not only for the understanding the individual effect of weather variables, joint effect of weather variables and interaction effect of the two variables but also very useful for students to learn about forecast model [12], Pre harvest forecast models for the rice and other crops as well. The work is also very important for researchers, for planners to prepare a plan on two and half months before harvesting of crop yield [13]. The work has importance on pricing and policy making, specially have the importance for implementation of agricultural development at village, block, District, State and country level as well.

References

Shankar U and Gupta BRD (1987) Forecasting the yield of paddy at chisurah in west bengalusing multiple regression technique.Mausam 38: 415-418.

Agrawal R, Chandrahas, Kumar A (2012) Used discriminant function analysis for forecasting crop yield. Mausam 63: 455-458.

http://www.rkmp.co.in/sites/default/files/ris/rice-state-wise/

Paramesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts across natural systems. Nature 421: 37-42.

LobellDB, BurkeMB (2010) On the use of statistical models to predict crop yield responses to climate change.Agric Forest Meteorol 150: 1443-1452.

Agrawal R, Jain RC, Jha MP (1986) Models for studying rice crop weather relationship.Mausam 37: 67-70.

Varshneya MC, Chinchorkar SS, Vaidya VB,Pandey V (2010) Forecasting model for seasonal rainfall for different regions of Gujarat. J Agromet12: 202-207.

Singh H, HundalSS, KaurP(2008) Effect of temperature and rainfall on wheat yield in south western region of Punjab. J Agromete10: 70-74.

Kumar N, Pisal RR, Shukla SP, Pandey KK (2014) Crop yield forecasting of paddy, sugarcane and wheat through linear regression technique for south Gujarat.Mausam 65: 361-364.

Pandey KK, Rai VN, Sisodia BVS (2014) Weather variables based rice yield forecasting models for faizabad district of eastern U.P.IntJrAgril& Stat Sci 10: 381-385.

MenzalA, Fabian P (1999) Growing season extended in europe. Nature 397: 659-663.

RafiZ, RehanA (2005) Wheat crop model based on water balance for agro-meteorological crop monitoring. Pakistan Jr Meteor 2: 23-33.

Root T L, Price J T, Hall K R, Schneider S H, Rosenzweig C and Pound J A (2003) Fingerprints of global warming on wild animals and plants. Nature 421: 57-60.

VEGETOS: An International Journal of Plant ResearchOnline ISSN: 2229-4473
Print ISSN: 0970-4078

Yield Forecasting Models based on Weather Parameters for Eastern U.P.

Abstract

Keywords: Correlation coefficient; MBE; PE; Rice; RMSE

Keywords

Introduction

Materials and Methods

Statistical Analysis of Data

Result and Discussion

References

Track Your Manuscript

H5 Index

Explore SciTechnol

Journal Highlights